Instructions to use tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO

SGLang

How to use tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO
```

Hyponatremia_L3_1000steps_1e7rate_03beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary4500_L3_100steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0025
Rewards/chosen: 0.4191
Rewards/rejected: -7.9725
Rewards/accuracies: 0.9980
Rewards/margins: 8.3916
Logps/rejected: -159.7724
Logps/chosen: -82.7927
Logits/rejected: -1.1012
Logits/chosen: -1.0642

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-07
train_batch_size: 1
eval_batch_size: 1
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6455	0.0112	50	0.6189	0.0168	-0.1483	0.7940	0.1651	-133.6916	-84.1338	-1.0991	-1.0690
0.1915	0.0224	100	0.2381	0.0998	-1.2728	0.9980	1.3727	-137.4402	-83.8570	-1.1007	-1.0693
0.0069	0.0336	150	0.0340	0.2244	-3.5445	0.9980	3.7690	-145.0125	-83.4417	-1.1014	-1.0678
0.0017	0.0448	200	0.0098	0.2714	-5.2540	0.9980	5.5254	-150.7106	-83.2852	-1.1013	-1.0664
0.0014	0.0559	250	0.0058	0.3321	-6.1233	0.9980	6.4554	-153.6084	-83.0827	-1.1013	-1.0655
0.0001	0.0671	300	0.0044	0.3409	-6.6530	0.9980	6.9939	-155.3742	-83.0536	-1.1000	-1.0641
0.0005	0.0783	350	0.0037	0.3524	-7.0398	0.9980	7.3922	-156.6634	-83.0152	-1.1004	-1.0643
0.0001	0.0895	400	0.0031	0.3703	-7.3960	0.9980	7.7663	-157.8508	-82.9556	-1.1006	-1.0643
0.0	0.1007	450	0.0029	0.4041	-7.5392	0.9980	7.9433	-158.3280	-82.8429	-1.1006	-1.0640
0.0	0.1119	500	0.0028	0.3938	-7.6566	0.9980	8.0503	-158.7193	-82.8773	-1.1011	-1.0644
0.0	0.1231	550	0.0027	0.3960	-7.7988	0.9980	8.1949	-159.1935	-82.8697	-1.1004	-1.0635
0.0001	0.1343	600	0.0026	0.4050	-7.8907	0.9980	8.2958	-159.4998	-82.8397	-1.1008	-1.0638
0.0	0.1454	650	0.0025	0.4102	-7.9529	0.9980	8.3630	-159.7068	-82.8226	-1.1006	-1.0637
0.0	0.1566	700	0.0025	0.4105	-7.9650	0.9980	8.3755	-159.7473	-82.8215	-1.1011	-1.0642
0.0037	0.1678	750	0.0025	0.4133	-7.9730	0.9980	8.3863	-159.7740	-82.8120	-1.1009	-1.0641
0.0	0.1790	800	0.0025	0.4059	-7.9812	0.9980	8.3871	-159.8014	-82.8367	-1.1012	-1.0644
0.0004	0.1902	850	0.0025	0.4003	-7.9906	0.9980	8.3909	-159.8326	-82.8553	-1.1015	-1.0645
0.0	0.2014	900	0.0025	0.4050	-7.9764	0.9980	8.3814	-159.7853	-82.8397	-1.1014	-1.0645
0.0	0.2126	950	0.0025	0.4187	-7.9726	0.9980	8.3913	-159.7726	-82.7940	-1.1012	-1.0642
0.0	0.2238	1000	0.0025	0.4191	-7.9725	0.9980	8.3916	-159.7724	-82.7927	-1.1012	-1.0642

Framework versions

Transformers 4.42.4
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/Summary4500_L3_1000steps_1e7rate_03beta_CSFTDPO

Base model

meta-llama/Meta-Llama-3-8B-Instruct

Finetuned

tsavage68/Summary4500_L3_100steps_1e6rate_SFT

Finetuned

(14)

this model